by Ophelia Deroy
Commentary on Mooijman, M., Hoover, J., Lin, Y., Ji, H. & Dehghani, M. Nat. Hum. Behav. https://doi.org/10.1038/s41562-018-0353-0 (2018)
Would the citizens of Rome have taken to the streets without Mark Antony’s famous speech against Brutus and the men who plotted against Cesar? What would have happened to the French revolution without Camille Desmoulins’ standing speech in the Palais Royal in 1789, or to the civil rights movement without Martin Luther King’s mastery of rhetoric in his ‘I have a dream’ intervention?
Playwrights and historians tell us of many famous speeches given at the onset of protests. But did words really play a causal role in pushing people to riot? Perhaps protests would have taken place anyway, speech or no speech, and we cannot run counterfactual history trials to know. The same question applies to the social media agitation that comes along with contemporary street protests: Words could play a causal role, but they could also have had a merely epiphenomenal moral outrage, accompanying but not influencing the protests.
Mooijman and colleagues tackle this tricky question about causality and use two sets of tools – Big data and controlled behavioural studies – that historians and political commentators did not previously have access to for deciding what stirred up street protests. They first focus on the tweets exchanged during the days of Spring of 2015, when protesters took the streets in Baltimore in response to the arrest and death of Freddie Gray in the hands of the US police. They use a machine learning algorithm to identify the moral expressions used in 18 millions of those tweets. “Moral rhetoric” in social media, they show, increased during the days of violent protests (volume of tweets also increased, so this means that the proportion of tweets with moral terms increases relative to other days).
Moral rhetoric here is limited to the explicit use of moral vocabulary, and more importantly, the observational method leaves open whether moral descriptions cause or correlate with protests. So what are we to infer? The stronger message here is that the authors show that the increase in moral tweets could also predict, hour by hour, the future counts of arrest during protests, which can be taken as a more convincing (albeit still indirect) and fine-grained measure of violence on the streets. Prediction, here, gets us nowhere closer to causal contribution, but at least moral outrage on social media can be seen as both diagnostic and prognostic of social movements.
The skeptic non-believers would argue that if both the tweets and the arrests were driven by a third factor, the predictive relationship would still be observed without requiring any causal relationship.
The paper then nicely digs into a special characteristic of social media: like a well-crafted speech, a tweet can highlight the moral significance of an event, which can turn into a reason to take the streets; it also tells how many people share or endorse this moral version of the events, by showing re-tweets and likes.
19yo has 500K bail for smashing a window on a cop car. #FreddieGray‘s killer is home on 350K bond. Even cop cars are worth more than a life.
— Cassandra Fairbanks (@CassandraRules) May 2, 2015
In controlled experiments, the authors provided people with more versus less heavily moralized descriptions of the same issue and asked them to say whether they would endorse the use of violence for such causes. They also modulated the same descriptions by adding information about the degree of agreement over these more versus less moralized issues. Both the degree of moralization and the represented degree of consensus played a role in people’s willingness to approve the use of violence.
What is trickier here is to know what the expressed agreement conveys. Does it help convincing people that a moral issue is at stake, then pushing them to accept more violence, or does it convince them only that more people would riot for such cause?
How informative is agreement is one of the key questions I have at present, and resonates with the present findings. Here, agreement may show that a given issue is a matter of shared moral value, rather than say, private preferences (which is what Goodwin and Darley (2012)[1]show for instance). Agreement may also give information about how many people, who are already convinced that the issue is a moral one, will be ready to go and join a protest or civil disobedience movement
[1]This paper shows that participants consider that the more consensual moral beliefs correspond to more objective moral values, while less consensual are seen as less objective. Negative values are also seen as more objective.
Recent Comments